AI News You Likely Haven't Seen Yet

Let's talk about some of the AI news that you might have missed from the past week. Let's start with GPT5. It recently came to everybody's attention that OpenAI has officially filed for a trademark for GPT5. You can see in the trademark they describe it as downloadable computer software for the artificial production of human speech and text. Downloadable computer software for natural language processing, generation, understanding and analysis. Supposedly it'll also include conversion of audio data files into text and voice and speech recognition. Now, OpenAI themselves haven't actually made any announcements about GPT5 or what they expect it to do yet, and there's been absolutely no indicator of when we can expect GPT5, but what we do know is that they did file for a trademark for it. In other ChatGPT news, OpenAI announced in a tweet that they were also adding some new updates to improve the Chat GPT experience. Things like prompt examples, suggested replies after you get an output. Now, if you've used something like Bing Chat, it will likely be similar to what we've seen in Bing Chat where you get a response and then it suggests some other potential questions or follow on conversation that you can have when you Open up chat GPT it actually defaults to GPT4 now if you're a plus user. If you're using code interpreter, they've added the ability to upload multiple files, which is apparently already working because I just uploaded two files and I just uploaded another. You'll now stay logged in so it's no longer going to log you out every two weeks. And they've also added some new keyboard shortcuts. I'm most excited about this ability to upload multiple images because we'll be able to do cool things like compare data from two spreadsheets, find correlations between two documents, or even upload three images and tell it to blend these three images into one, which when I gave it that prompt, it ran some code and generated this image of the three different files I gave it, where it just sort of overlaid them over each other. So this multi file ability is a really exciting new upgrade in my opinion. Speaking of ChatGPT, a few weeks ago if you had gone to AI.com, it would have redirected you to ChatGPT. However, now if you try to go to AI.com, it actually redirects you to X AI. So it appears that opening OpenAI used to own AI.com, but now Elon and XAI own AI.com so I'm actually not certain. And if I read this article on TechCrunch, they're really not certain either whether or not elon always own AI.com and used to point it to chatgpt and now moved it to X. Or if Elon somehow bought it off of OpenAI, which I can't imagine OpenAI would have sold it to Elon, but who knows? It sounds like there's a lot of speculation. I don't know if Elon or anybody at OpenAI is going to just come out and say why AI.com is now going to X AI. Just something interesting to note. Also this week YouTube started testing AI generated video summaries. You can see if you head over to YouTube's help document. They actually have a section for YouTube test features and experiments and on July 31st they said they were testing AI auto generated summaries. We're starting to test AI auto generated summaries on YouTube so that it's easier for you to read a quick summary about a video and decide whether it's the right fit for you. We don't know much more than that yet, but but I highly doubt the summaries are gonna be in depth enough that they're gonna make you not watch the video. I believe they're meant to be helpful in making the decision to watch a video but not replace actually watching the video. Also recently Stack Overflow announced Overflow AI. If you're not familiar with Stack Overflow, it's probably because you're not a coder. If you are a coder, you're probably very familiar with Stack Overflow because this is where coders go to get help when they're trying to generate code and they run into errors, or if they're trying to find code that somebody else has already written and find solutions to problems that they're having. Coding Process According to this article on similarweb.com stack overflows traffic has steadily been decreasing since ChatGPT has come out and since GitHub released their AI copilot. Because now coders are just turning to tools like Chat GPT and GitHub Copilot when they run into issues with coding and they don't need to go to Stack Overflow to find where somebody else had already solved the problem. So now they're trying to get the traffic back to their site by introducing Overflow AI, you can get generative AI summarized answers based on the combined results of Stack Overflow's knowledge base of 58 million questions and answers. So basically all of the questions and answers that have been put on Stack Overflow over the years are going to be much more easily searchable through Overflow AI. And also Overflow AI promises to cite its sources, so wherever it's pulling code from inside of its database, it's going to tell you where it came from. It's also going to be interesting to see how all of the contributors to Stack Overflow over the years feel about the fact that now all of the information they've inputted is now being trained into this AI. Because as we've seen in the AI world so far, there's a lot of pushback when explicit permission wasn't given to train on data that other people created. Speaking of that, Kickstarter just announced new requirements and disclosures for generative AI projects. Projects will now be required to disclose relevant details on their project pages going forward. These details must include information about how the project owner plans to use the AI content in their work, as well as which components of their project will be wholly original and which elements will be created using AI tools. Projects involved involving the development of AI tech tools and software must detail info about the sources of training data the project owner intends to use. In similar news, Instagram is working on labels for AI generated content. Now, we don't know a ton about this yet, but here's a sort of leaked screenshot of something that's going on behind the scenes at Instagram. According to Alessandro Pelusi, Instagram is working to label the contents created or modified by AI in order to be identified more easily. And here's a screenshot of what this this little warning looks like. The creator or meta said that this content was created or edited with AI. The image was generated by Meta AI, and then it gives an explanation of what generative AI is. And then it has a disclaimer how to know when posts use AI. Content created with AI is typically labeled so that it can be easily identified. This is really all anybody is going off of right now. But what we can gather from this it looks like the creator will have the option to claim it was made by AI, but it also looks like Meta themselves will somehow be trying to determine whether something was generated with AI. Most likely, if it was generated with Meta AI, Meta will be doing their own sort of watermarking of the image. So Meta knows if it was generated with Meta's AI, but how it's actually going to know if it was generated with some other AI tool outside of Meta's ecosystem yet to be seen. Speaking of Meta and their AI. This week they introduced AudioCraft, a generative AI tool for audio and music. They released a news article about it over on their facebook.com news page and they also released some more research and some more samples of what it can do on the Meta blog. So if we look at the key takeaways here, Audiocraft lets you easily generate high quality audio and music from text, and they claim they're open sourcing music Gen Audiogen and Incodec, the three tools that sort of make up this Audiocraft platform. So checking out some of the samples over on the Meta blog here. Here's some examples of generating audio from text descriptions. Here's the prompt Whistling with wind blowing.

Here's another prom sirens and humming engine approaches a pass.

Now that is AI generated that sounds like it could have just been ambient noise recorded out of a window in a city. And then their Audio gen actually generates music from text prompts. So pop dance track with catchy melodies, tropical percussions and upbeat rhythms. Perfect for the beach.

Feel like I'm playing Monkey Island. Here's another example. Text prompt, earthy tones, environmentally conscious, ukulele infused harmonic, breezy, easygoing, organic instrumentation, gentle grooves.

I mean it's pretty dang good what these tools can do. And if we scroll down to the bottom of the blog here, they actually have a link to browse Audiocraft code which links off to their GitHub page where you can actually access the code. Right now it includes music gen Audiogen and their incod. It looks pretty dang open source to me. The license gives access to commercial use, modification, distribution and private use, and doesn't seem to have the same limitations that they put on llama2. Now speaking of llama2, the company Alibaba is actually open sourcing their AI model. I don't really want to try to pronounce what their model is called, but it's something along the lines of Tong Yi Qianwen. They released this model back in April and it actually did allow content generation in both English and Chinese. However, the announcement that came this week is that Alibaba is going to be open sourcing the 7 billion parameter model called Quinn 7B, as well as a version designed for conversational apps called Quin 7B Chat. So it's basically going head to head with what Meta has released in their open source llama 2. Next, I want to tell you about a really cool video creation tool called Invideo, who happens to be the sponsor of today's video. Invideo makes it incredibly easy for anybody to make professional, professional quality videos in minutes. They have a huge library of video templates, an absolutely massive library of stock media and a super intuitive drag and drop editor. So I could literally do all of my video editing straight from within their platform if I wanted to. But let's check out their cool AI text to video feature here. So first we can browse their template library to find one that fits my needs. Let's go with this futuristic technology facts, you can see that it already pre built an entire video that I can just swap out the text inside of the video. I think this template worked well and fits in with the type of content I make on this channel. So let's go ahead and use this template and I could either copy and paste a script in here or I can ask AI to write a script for me. So let's describe our script how generative AI and robotics will change the world and click generate. And it generated a script for us. So let's go ahead and click generate scenes. And you can see it generated a whole bunch of scenes for each little chunk of my script here. And let's say one of the scenes here doesn't quite fit what the text says that the AI generated. It's really easy to swap out the scene. One of the best parts about Invideo is their stock media library. With a premium account you get access to over 8 million royalty free videos, images and audio tracks from top providers like iStock, photo and Shutterstock literally valued at hundreds of dollars. Let's go ahead and search out shopping clips and drag this media right here. Click add. We'll go ahead and delete this original media here. And now our media lines up with what it's saying a little bit better and with a premium plan can now download my video in full 1080p with absolutely no watermark on it. This is great for making things like explainer videos, faceless YouTube channel videos, product demos and so much more. And they Even have live 24 hour chat support right inside of the platform. And right now you can actually get 25% off if you use the coupon code MW25 at checkout when signing up for a premium Invideo account. I'll make sure to drop the link in the description, but don't forget to use that coupon code MW25. InVideo also has a really cool new platform coming called Idea to Video. Like make a video on how to live inside a black hole without dying in rhyme in the voice of Morgan Freeman, Click Generate and it will actually make that video for you. A giant bubble wrap suit you'll see will protect you from gravity's glee. And with that one text prompt, it will create a video with the script, with the text, with the voiceover, and with the style footage overlaid on the video, all from a single text prompt. And if you have a premium Invideo account, you're just going to get access to this when it's available. And finally, I can't talk about Invideo without talking about their really cool YouTube channel where they give you all sorts of really cool ideas on how you can use the Invideo platform to further your business or improve your YouTube channel. Their channel is called Invideo Official and once again I will make sure it is linked up in the description below. Thanks so much again to Invideo for sponsoring this video. And if you are going to go grab Invideo and make sure you use the coupon code MW25 for 25% off a premium account. Also this week Google announced three new features of their generative AI search engine. Now I don't believe everybody has access to the generative AI search engine yet, but I actually popped over to my Google account today and noticed a new button down here that says try AI Powered Overviews when you search. I've never seen this before. I don't know if this is fully rolled out or if I have a beta account, but I can actually select a pre selected prompt and enter their search experience right now. For example, let's search how do Penguins stay Warm? And you can see I get this generative AI experience up here. Now the new features they rolled out are get a better understanding through images and video. So when we search we notice we actually get some images within our search as well as some video suggestions. But I believe this might just be part of the regular search engine here and we get a chat response with some external links here it says get your AI powered overviews faster. And they recently added more information to the responses like the published dates of the links. Let's try another prompt here. Benefits of Buying Secondhand Clothing Here are some benefits of buying secondhand clothing and it looks like it just puts the chat right above the regular search results that you get anyway and then it gives you some links to the articles that it's referencing. So kind of similar to what we'd see in something like Bing chat right now. So here's something interesting. Nvidia developed a new AI method called perfusion. This method only takes up 100 kilobytes of space, which is small enough to fit on an old school floppy disk and of course is way smaller than models like Stable Diffusion or Midjourney. It can also train in just four minutes. You can see here are some examples of some generations that it did. A cat is acting in a play, wearing a costume. This is the version that this teeny tiny model generated versus what Custom Diffusion and Dream Booth generated. A teddy is playing with a ball in the water. Here's the versions they generated. They introduce a new idea called key locking and they describe it this way. It works by connecting new concepts that a user wants to add, like a specific cat or chair to a general category. During image generation, for example, the cat would be linked to the broader idea of a feline. This helps avoid overfitting, which is when the model gets too narrowly tuned to the exact training examples. By tying the new cat to the general notion of a feline, the model can portray the cat in many different poses, appearances and surroundings, but it still retains the essential catness that makes it look like the intended cat, not just any random feline. Now, if I'm being totally honest, I'm still a little bit confused about how this model works myself and how they managed to get it so small because, I mean, that's smaller than most image files that you would generate these days. But I'm excited to learn more as more information comes out about it and as it actually becomes available for us to use. Nvidia has only presented the research paper for now, promising to release the code soon, so we'll just have to wait and see. But what we do have the code to right now is one of my favorite AI image generation tools called Leonardo. And they just rolled out a brand new Photo Real feature along with a few minor user interface updates. You you can see there's a new toggle here when you log into Leonardo for Photo Real. And if we turn on Photo Real, we actually have a few options here of turning the depth of field to low, medium or high, and turning RAW mode on or off. They recommend leaving RAW mode off because it requires more detail in your actual prompting. Now, I made a video a couple weeks ago about Leonardo and why I love Leonardo so much. And I also mentioned that they're about to release a photorealism update and it is here. I just used this prompt. A man in a colorful shirt standing in front of a camera and holding a microphone. News Reporter. Cartoony octane render. This is one of the prompts that I like to use to generate thumbnails of these exact news videos. And then I actually go in and replace the face with my own face. And you can see the photorealism here is pretty impressive. You can see there's a few options here for cinematic, creative and vibrant. Let's go ahead and try the same prompt here using Vibrant and look what it generated. The details are really good. I mean, this almost looks like Ken from the Barbie movie, but I like it. I might actually use one of these images as my thumbnail for this video. You can see last time I used this prompt before the Photo Real version was available, this is what it came up with. Quite a bit improved, I believe. The photoreal option is only available for premium users with a paid account, but they also have free trials right now on the paid account. This week, InWorld AI became the best funded startup in AI and gaming. If you remember, I did a video a few weeks back about a tool called Convey, which as AI personality and large language model speech generation to game characters. InWorld AI is a very similar concept. It's a tool where you can have game characters that have fully AI generated conversations with you. They've raised over $120 million since it was founded and it now has a valuation of a half a billion dollars. Now when I really look at all of the AI stuff that's coming out these days, I feel like there's a few areas where AI is really just going to change everything. One of them being healthcare. I think we're just going to see dramatic changes in the world of healthcare. But the second is in the gaming world. I really see AI having huge impacts on the gaming world. Apparently investors are seeing the same thing because they're dumping tons of money into companies like InWorld, Convey and even that common sense machines tool that I was showing in a previous video. Now the other area where I think we're going to see AI have a dramatic impact on is, well, military and weapons, which is also kind of scary. Here's a news article about how the US Air Force recently completed a successful test flight of the XQ58 Valkyrie drone using AI control. The test demonstrated the drone's ability to perform air combat tasks anonymously using AI. The algorithms were trained extensively on simulation data from all sorts of test runs done in a simulator. This allowed the military to train these drones very cheaply without the need to actually train on real flight data. We also know that other countries are out there using AI to train autonomous military vehicles. So we may see a future where we're just creating robots to go fight other robots that are all trained with AI. So maybe the future of warfare could be our country's billion dollar equipment shooting down other countries, billion dollar equipment and vice versa. But it's all unmanned. I don't know. This is one of the scarier areas of AI where I'm not sure. I totally want to see how it all plays out. And before I wrap up, I want to talk really quickly about this sort of subsection of the Internet, which I find absolutely fascinating. I'm not sure if you're familiar with Lil Miquela. She's got 2.7 million followers on Instagram, 279,000 subscribers on YouTube, 30,000 followers on Twitter, and 3.5 million followers on TikTok. Now, the thing about Lila Michaela is that she's not a real person. This is all AI generated. This is an AI influencer. And this is becoming a very interesting and hot topic right now because this specific AI influencer charges $10,000 per Instagram post and has an eight figure net worth. She's an AI character with an eight figure net worth. That is just mind blowing to me. And the only reason this actually came to my attention is I was in a hotel room a couple days ago watching tv and a commercial came on and I actually saw Lil Miquela in a Calvin Klein ad on tv. And I went, I don't think that was a real person. I think that was AI generated. And I started digging and I found out more about Lil Miquela. I mean, I'd heard the name months and months and months ago, but never really put too much thought into it. And the fact that this AI influencer was making as much money as she, it was actually generating as much money as they're generating. So that was just wild to me. I wanted to point that out because I went down this rabbit hole a little bit and I figured maybe a few of the people that are watching this video would want to go down that rabbit hole as well. And before we wrap up, I want to point out this really cool video from this YouTube channel that I called AI Fiction Factory, where they actually created a video called Sending Matt Wolf to Planet X featuring elon Musk. An AI trailer. And they used all AI generation tools such as Midjourney Runway, ML, Pika Labs, Claude 2Eleven Labs, Full Journey and more, to generate this minute long trailer for a video where they send me to the Planet X. So I wanted to give a little clip of it and I'll link up to the YouTube channel below if you wanna watch the full video. Video.

My name is Matt Wolf. And I'm an interstellar ambassador from planet Earth. I've been assigned to report on life on Planet X. Little did I know just how different things would be. That's as far as I'm gonna go. I don't want to steal their thunder. I want you to go watch the video on their YouTube channel. You can find it over at AI Fiction Factory. And I just want to thank them so much for making that video. I just think it's awesome that there's people out there that are actually making. Making content around me with my face and my voice in it. It's definitely interesting and a different place than I was in a couple years ago where nobody knew who the heck I was. I was just some random dude making content on the Internet about cool stuff that nobody found yet. And now I've got people sending me to Planet X. So thanks for doing that. I think it's really awesome. I love seeing stuff like this and how other people are pulling these tools together to create really interesting stuff. So that's all I got for you this week. It wasn't a major News Week in AI this week. However, next week is the 6Siggraph Conference, a conference where all sorts of researchers come out and show off a lot of the work they've been doing. And some of the speakers at the Siggraph concert include Dario Gill from IBM Research, Jensen Huang from Nvidia, who people are expecting might make some pretty interesting announcements at the event, and many other big names. I'll be at this conference most of the week, but the point I'm trying to make here is that there's probably going to be quite a few announcements next week because. Because I have a feeling a few of these companies are holding out for Siggraph to make some of the big announcements, which is why this week might have been a little bit slower of a week. So excited to see what comes out at Siggraph, what sort of research is unveiled. I'm going to try to bring my cameras in. I do have a media pass, so I'm going to be walking around trying to shoot footage and capture as much as I can so that I can come home and make a recap of some of my biggest takeaways from Siggraph. It should be an exciting week. My video posting schedule might be off a little bit bit for the month of August because I am going to a handful of conferences. By the end of August, I'll be back on my normal schedule cranking out cool AI videos for you. And I'M going to keep on making these end of the week AI News videos for you to keep you in the loop with everything that's been announced. Once again, I appreciate you tuning into this channel and checking out these videos. If you haven't already, check out Future Tools IO. I keep the AI News page up to date pretty much every single day. If you go to the homepage you can find all of the cool AI tools that I curate on almost a daily basis. And if all this is too overwhelming, there's a little too much for you. Join the free newsletter and I'll just give you the TLDR in your email once a week. Pretty soon gonna be twice a week, but as of right now it's once a week. It's the TLDR of AI for the week. I'll keep you in the loop. It's totally free to join and you can find it at FutureTools IE and click Join the free Newsletter thanks again to Invideo for sponsoring this video. I really, really appreciate it. I appreciate you for watching. Thank you so much. I'll see you in the next video. Video Bye bye.